Plasmonic Ferroelectric Modulators

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 37, NO. 2, JANUARY 15, 2019 281 Plasmonic Ferroelectric Modulators Andreas Messner , Felix Eltes ,PingMa ,StefanAbel , Benedikt Baeuerle , Arne Josten , Wolfgang Heni , Daniele Caimi, Jean Fompeyrine , and Juerg Leuthold , Fellow, IEEE, Fellow, OSA

(Invited Paper)

Abstract—Integrated ferroelectric plasmonic modulators fea- travelling wave electrodes suffer from high electrical losses at turing large bandwidths, broad optical operation range, resilience high frequencies and the walk-off between electrical and opti- to high temperature and ultracompact footprint are introduced. cal signals [6], both of which ultimately limit the modulation Measurements show a modulation bandwidth of 70 GHz and a temperature stability up to 250 °C. Mach–Zehnder interferome- bandwidth and power efficiency of the devices. ter modulators with 10-µm-long phase shifters were operated at The need of higher integration densities led to the emerging of 116 Gbit/s PAM-4 and 72 Gbit/s NRZ. Wide and open eye diagrams the silicon (Si) photonics platform. Si photonics was proposed with extinction ratios beyond 15 dB were found. The fast and robust in 1985 [3]. It relies on silicon-on-insulator (SOI) wafers and devices are apt to an employment in industrial environments. promises to leverage scaling effects similar to the mature CMOS Index Terms—Electrooptic modulators, ferroelectric devices, technology and more compact footprint due to large contrasts high-speed integrated circuits, plasmonics. in refractive indices between adjacent layers. However, Si does I. INTRODUCTION not exhibit the Pockels effect so that Si modulators commonly resort to the plasma dispersion effect. This effect faces a trade- IGH-SPEED, compact and power-efficient electro-optic off between modulation speed and strength and acts on both H (EO) modulators are currently in the spotlight of phase and amplitude of the optical signal. Therefore, a func- research as they are key components in high-capacity optical tional material which exhibits the Pockels effect and which can links. Many physical effects have already been exploited to be cointegrated with the Si photonics platform is of great interest perform electro-optic (EO) modulation. Among them are for for the development of high performance EO modulators. example, the quantum-confined Stark effect [1], [2], the plasma A possible research direction towards this aim is the silicon- dispersion effect [3], [4] or the linear EO effect, commonly organic-hybrid (SOH) platform [7], [8]. There, functional or- called Pockels effect. Here, the Pockels effect is of particular ganic materials offering the Pockels effect are introduced and interest as it provides a large optical bandwidth and a pure modulators have already demonstrated bandwidths of 100 GHz phase modulation such as needed to operate phase shifters [9] and data rates of up to 400 Gbit/s [10]. in MZ and IQ modulators. Both MZ and IQ-modulators are Another technology is the “lithium niobate on insulator” key elements for encoding advanced modulation formats in (LNOI) platform, which is based on a thin film of LNB on high-capacity communication systems. a thick insulating SiO2 layer. Being proposed in 2010 [11], it State-of-the-art EO Pockels modulators commonly rely on has attracted an increasing attention [12]Ð[14]. Due to the suf- 3 the Pockels effect of lithium niobate (LiNbO ,LNB),aferro- ficient refractive index contrast between LNB (nLNB ≈ 2.2 at electric material. Transmission rates of 1.6 Tbit/s have already λ = ≈ 1550 nm) and SiO2 (nSiO2 1.44), this platform omits been demonstrated using a LNB IQ modulator [5]. Yet, those Si completely and still allows miniaturization of basic passive LNB modulators are typically based on weakly guiding waveg- photonic components, such as ridge waveguides, Y-splitters, uides, resulting in centimeter-long devices [6]. Also, the long multimode interference couplers (MMIs), and resonators [12], [15], [16]. Active components such as EO modulators are di- Manuscript received August 4, 2018; revised October 15, 2018; accepted November 7, 2018. Date of publication November 14, 2018; date of current ver- rectly enabled by the Pockels effect of LNB and benefit from sion February 20, 2019. This work was supported in part by Swiss National the platform’s strong modal confinement and effective overlap Foundation under Project 200021_159565 PADOMO and Project IZCJZ0- between optical and electric signals [12]. Additionally, the low- 158197/1 FF-Photon, in part by the European Commission under Grants FP7- ICT-2013-11-619456-SITOGA, H2020-ICT-2015-25-688579 PHRESCO and permittivity SiO2 layer facilitates velocity matching between H2020-ICT-2017-1-780997 plaCMOS, and in part the Swiss State Secretariat the two signals to further increase the modulation bandwidth. for Education, Research and Innovation under contracts 15.0285 and 16.0001. Recently, Zhang et al. reported a 100 GHz bandwidth Mach- (Corresponding author: Ping Ma.) A. Messner, P. Ma, B. Baeuerle, A. Josten, W. Heni, and J. Leuthold are with Zehnder modulator (MZM) with 5 mm long phase shifters and the Institute of Electromagnetic Fields, ETH Zurich,¬ Zurich¬ 8092, Switzerland a voltage length product of Vπ L = 2.2 Vcm [14]. While the (e-mail:,[email protected]; [email protected]; [email protected]; ajosten@ethz. LNB platform has made significant progress beyond state-of- ch; [email protected]; [email protected]). F. Eltes, S. Abel, D. Caimi, and J. Fompeyrine are with IBM ResearchÐ the-art, one of the main limitations of the technology is the small Zurich, Ruschlikon¬ 8803, Switzerland (e-mail:, [email protected]; sab@ substrate size (<6 inch), which is incompatible with current zurich.ibm.com; [email protected]; [email protected]). CMOS standards. In addition, advanced device designs featur- Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. ing smaller footprint, co-integration with Si electronics and even Digital Object Identifier 10.1109/JLT.2018.2881332 lower power consumption deserve more research efforts. 0733-8724 © 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications standards/publications/rights/index.html for more information. 282 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 37, NO. 2, JANUARY 15, 2019

Fig. 1. False-color SEM images of (a) Plasmonic ferroelectric Mach-Zehnder modulator; (b) a plasmonic phase modulator. (c) Close-up of a tapered mode converter. The Si bus waveguide and tapered mode converter are used to couple the photonic mode into the Au-BTO-Au plasmonic slot waveguide. Insets: Simulated photonic and plasmonic mode profiles, respectively. (d) Simulated electrical field of the photonic mode in the Si waveguide (TE mode). (e) Simulated electrical field of the plasmonic mode (transverse component). (f) Simulated RF field in the slot waveguide (transverse component). We assume r,BTO = 1000.

And indeed, there are two opportunities to further reduce [33]. While these demonstrations are very successful, organic the voltage-length product and simultaneously increase the EO materials encounter reservations by the industry as a CMOS bandwidth: The first is to choose a material with a higher Pock- compatible solution of thermally stable materials for low-cost els coefficient than LNB and the second is to switch to a plas- mass-produced transmitter products. monic device geometry which could provide higher modulation In this paper, we report on the current progress in realizing efficiency and smaller RC time constant than typical photonic ultra-compact and high-speed plasmonic ferroelectric modu- devices. lators (PFMs) by effectively combining the epitaxially grown The first opportunity to decrease the voltage length product functional ferroelectric BTO material with a plasmonic device arises from switching from LNB with a Pockels coefficient of design. We present 72 Gbit/s non-return-to-zero (NRZ) data only 30 pm/V [17] to other ferroelectric materials that exhibit modulation and 116 Gbit/s 4-level pulse-amplitude modulation a stronger Pockels effect. Barium titanate (BaTiO3 ,BTO)is (PAM-4) experiments. Experiments are performed with modu- among the most promising ferroelectric materials for EO ap- lators featuring an active section that is as short as 10 μm, a plications, because of its strong Pockels effect [18]. Its largest measured frequency response that exceeds 70 GHz (only lim- Pockels tensor element is reported to be r42 = 1300 pm/Vin ited by the experimental instruments) and a record low voltage- the unclamped, zero-stress case, and to be r42 = 700 pm/Vin length product between 150 and 200 Vμm, depending on the the clamped case [19]. The Pockels effect of BTO has already modulation frequency. Furthermore, temperature stability up to been exploited in active Si photonic devices [20]Ð[24] and [53], 250 °Cisshown. since BTO can be epitaxially grown on Si [20]. Recently, EO The paper is organized as follows. Section II describes the modulators based on BTO have even been monolithically inte- device design and concept. Section III depicts the fabrication grated on an advanced Si photonic platform, opening up possi- process of the device. Section IV discusses the influence of bilities towards monolithic integration with electric circuits in a BTO’s crystalline orientation on the device design. A deriva- foundry environment [24]. tion of the effective Pockels coefficients is given. Section V The second opportunity to decrease the voltage-length prod- presents passive as well as EO characterization experiments. uct and simultaneously raise the EO bandwidth dramatically is These results include the frequency response of PFMs and the to transit from photonics to plasmonics. A plasmonic modulator temperature stability test. Section VI shows the data modula- device can be built using a metal-insulator-metal slot waveguide tion experiments of 116 Gbit/s PAM-4 and 72 Gbit/s NRZ. geometry. Here, metal electrodes are used both as guiding struc- Section VII concludes the work by summarizing the key results. tures for the optical mode and as radio frequency (RF) electrodes for the electrical signal. The slots can be filled with nonlinear II. DEVICE CONCEPT AND DESIGN functional materials introducing the Pockels effect. This leads to This section presents the concept and design of the efficient, the so-called plasmonic-organic hybrid (POH) approach, which compact, and ultra-fast modulators. Two plasmonic ferroelec- adopts nonlinear organic materials filled into a narrow, plas- tric phase shifters are integrated into a photonic Mach-Zehnder monic metallic slot waveguide. The POH technology offers en- interferometer (MZI). Photonic elements are employed for fiber- hanced modulation efficiency and bandwidth [25], [26]: To this to-chip coupling and on-chip routing. point successful high-speed modulation has been demonstrated onafewμm2-footprints [27]Ð[29] with high-speed data modulation of up to 200 Gbit/s [30] that can be related to a flat A. Plasmonic Ferroelectric Mach-Zehnder Modulators EO frequency response up to 325 GHz [31], [32] and a record Fig. 1(a) shows a fabricated PFM, which consists of two fer- low power consumption of as little as 2.8 fJ/bit at 100 Gbit/s roelectric phase shifters, Fig. 1(b), in a MZI. We employ Si MESSNER et al.: PLASMONIC FERROELECTRIC MODULATORS 283

discussed here, the field interaction factor is typically larger than 1, which means that the material response is actually amplified by the waveguiding structure. Low drive voltages Ð accompanied by low power consumption Ð can as well be traced back to a few geometrical advantages of the concept. For instance the small wavelength of plasmons allows for the guiding in narrow slot waveguides. This then provides another advantage. Thanks to these narrow slots (width: d), one obtains a very strong electrical field E = V/d, which is Fig. 2. A cross section through the plasmonic ferroelectric phase shifter. 80 nm the key stimulus for the Pockels effect, described by the Pockels BTO is wafer-bonded to 3 μmSiO2 on Si, using a 20-nm-thick Al2 O3 buffer layer. Two gold electrodes enclose a narrow (50 ...150 nm) and 40-nm-tall coefficient r. BTO slab. V 1 ΔnBTO ∝ r · E = r · ∝ (2) d d photonic grating couplers [34] to couple the light to waveg- Voltage requirements to operate a PFM are further relaxed uides on the chip and multimode interference couplers (MMIs) by the PFM’s open-circuit behavior and the push-pull configu- to split up the light into the two arms of the PFM. A tapered ration of the MZM. Because their parasitic impedance can be mode converter, Fig. 1(c), couples light from the 450-nm-wide neglected with respect to 50 Ω driving electronics, the devices Si access waveguide, Fig. 1(d), to the plasmonic mode, Fig. 1(e), appear as open-circuit terminated, and reflect the incoming RF and at the end of the phase shifter the plasmonic mode is cou- signal. This leads toa6dBvoltagegainascomparedtoa pled back to the Si waveguide by another mode converter. Both terminated device with travelling wave electrodes. Drivers that the photonic and the plasmonic mode are TE-polarized. The withstand the reflected wave are commercially available and asymmetry in the MZI arm lengths (100 μm difference) allows have been successfully used in our experiment. Another factor for an easily accessible evaluation of the fabricated devices and two enhancement in voltage is gained from joining two plas- supplements the capability of the voltage-controllable operating monic phase shifters to a MZM. The nature of the linear EO point by offering an additional spectral tunability. effect allows for operating the MZM in a push-pull configu- Plasmonic ferroelectric phase modulators are the key compo- ration: Phase modulations of opposite signs are generated in nents of our devices. They are based on an Au-BTO-Au plas- the two arms and added up at the output. While this halves monic slot waveguides, a concept which has already been suc- the needed driving voltage, it also results in a pure, inherently cessfully demonstrated with organic EO materials [25], [35]. A chirp-free modulation. schematic cross-section of such a phase modulator is shown in Thanks to this high efficiency, a single phase shifter is as ( ) Fig. 2. The metal electrodes form a very narrow 50 ...150 nm short as 10 μm and does not require travelling wave electrodes. BTO-filled slot, and act both as optical waveguides and as feeds Its footprint can be as small as 20 μm2, although it is currently for the RF field. constrained by the size of the electric contact pads. Plasmonic modulators offer ultra-broad EO bandwidths: The B. Characteristics of Plasmonic Ferroelectric Modulators compactness dramatically reduces the devices’ parasitic capac- PFMs feature efficient modulation, low drive voltages, a most itance C. Due to the metallic RF feeds, the devices also feature compact footprint while offering broadband frequency response. a very small parasitic resistance R. As a result, plasmonic slot The PFMs are efficient because they benefit from three dis- modulators offer a uniquely small RC time constant, which en- tinct advantages over more conventional modulators. First, they ables EO bandwidths pushing to the THz regime [26], [39]. In benefit from BTO as an EO material, which provides one of the strong contrast to carrier effects, the Pockels effect is an ultra- highest Pockels coefficients among all known materials [18]. fast field effect and does not jeopardize the device’s frequency Second, they benefit from an almost perfect overlap between the response [40]. EO material, the optical mode (nBTO ≈ 2.27 at λ = 1550 nm ABRICATION ROCESS [36]) and the electrical field, (r,BTO ≈ 1000, measured at III. F P 20 GHz [37], [38]), as shown in Fig. 1(e) and (f), respectively. The devices discussed here are fabricated on a Si/3 μm Third, a high group index leads to field enhancement due to SiO2 /20 nm Al2 O3 /80 nm BTO/220 nm Si wafer. To man- a slow-down effect. All of these effects strongly change the ufacture this layer stack, an 80-nm-thick BTO film was de- effective refractive index neff of the plasmonic mode, posited epitaxially on a SrTiO3 -coated SOI wafer by molecular beam epitaxy (MBE). The 4-nm-thin SrTiO3 layer is neces- Δneff =ΔnBTO · Γ (1) sary to enable an epitaxial relationship between Si (100) and Here, ΔnBTO is BTO’s refractive index change, which is BTO [20]. Using 10-nm-thick Al2 O3 adhesion layers, the wafer proportional to its Pockels coefficient, and Γ is the field inter- was bonded [41] to a Si wafer with 3-μm-thick thermal SiO2 action factor [7], [8]. The Γ-factor does not only comprise the cladding. The original BTO handle wafer and its buried oxide strong field confinement and overlap, but additionally reflects (BOX) layer were stripped. The resulting layer stack resembles the slow-down effect, because it is inversely proportional to the a standard SOI wafer with a functionalized BOX layer directly Γ ∝ −1 group velocity vg , ng [7]. In plasmonic structures such as adjacent to the device Si layer. 284 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 37, NO. 2, JANUARY 15, 2019

TABLE I BTO POCKELS TENSOR COMPONENTS

by using the tensorial expression as follows ⎛ 2 ⎞ ⎛ ⎞ Δ 1/n 00r13 1 ⎜ 2 ⎟ ⎜ ⎟ ⎜Δ 1/n ⎟ ⎜ 00r23⎟ ⎛ ⎞ ⎜ 2 ⎟ ⎜ ⎟ EΩ Fig. 3. (a) Crystal structure of the BTO unit cell in the tetragonal phase. ⎜Δ 1 2 ⎟ ⎜ 00 ⎟ x ⎜ /n 3 ⎟ ⎜ r33⎟ ⎜ ⎟ The a-andb-axis have the same length, the c-axis is slightly longer. The ⎜ ⎟ = ⎜ ⎟ · ⎝EΩ ⎠ (3) titanium atom resides in a double-well potential whose minima are slightly ⎜Δ 1 2 ⎟ ⎜ 0 0 ⎟ y ⎜ /n 4 ⎟ ⎜ r42 ⎟ Ω shifted from the volume center along the c-axis. This gives rise to BTO’s ⎜ ⎟ ⎜ ⎟ E spontaneous polarization, ferroelectricity and its Pockels effect. (b) Orientation Δ 1 2 00 z ⎝ /n 5 ⎠ ⎝r51 ⎠ of waveguide within BTO crystal as discussed in this work. The waveguide lies Δ 1 2 000 in the (ac)-plane. Its direction is offset by an angle ϕ with respect to the c-axis. /n 6 The RF field EΩ and the optical polarization are oriented transversely to the waveguide direction. Ω E is the applied external field, rmn are the Pockels coefficients in the Voigt notation [43] (x =1,y =2,z =3;xx =1,yy= 2,zz=3,yz =4,xz =5,xy =6). The crystallographic point E-beam lithography (EBL) was used for all resist exposures. group 4 mm of BTO explains the zero entries of that matrix. = = The Si photonic components were structured by a chemical Moreover, it is r13 r23 and r42 r51. Values of the Pock- dry etching process. Subsequently, 50-nm-wide and 40-nm-tall els tensor components reported by Zgonik et al. are given in BTO strips were structured by a physical dry etching process. Table I [19]. The Au electrodes were applied by e-beam evaporation and a The r42 and r51 components are by far the strongest tensor lift-off process [23]. elements. The question arises if the EO response can be maximized by changing the angle ϕ between the waveguide and the c-axis, see Fig. 3(b). To this end, we calculate the effective Pockels IV. EFFECTS OF CRYSTAL STRUCTURE ON POCKELS EFFECT coefficient reff in dependence on ϕ and follow the derivation The crystalline structure of BTO determines the nonlinear given in ref. [44]. EO response of the material. In this section, we first discuss We start by stating the index ellipsoid of BTO, which is the EO properties of the bulk single crystal and calculate the equivalent to the tensorial notation of Eq. (3). Furthermore, we effective Pockels coefficient related to an optical wave. Second, use that r13 = r23 and r42 = r51: we calculate the effective Pockels coefficient of the MBE-grown 1 1 1 + 2 + + 2 + + 2 BTO thin-film considering its multi-domain nature and compare 2 r13Ez x 2 r13Ez y 2 r33Ez z the extracted coefficients to the values reported in the literature. no no ne +(r42Ey )2yz +(r42Ex )2zx =1. (4) The x-, y- and z-axes correspond to crystal’s a-, b- and c-axes, A. The BTO Bulk Crystal see Fig. 3(a). We assume the waveguide to be in the xz-plane, Fig. 3(a) shows the unit cell of BTO. At room temperature, as indicated in Fig. 3(b). The coordinates x and z give the BTO crystallizes in a tetragonal, pseudo-Perovskite structure, waveguide’s transversal and longitudinal direction, respectively. where the a- and b-axis of the crystal have the same length, The coordinate transformation between (x, z) and (x,z) is while the c-axis is slightly longer. In bulk, the lattice constants given by = = Aû = Aû are a b 3.991 and c 4.035 [42]. The titanium atom Aû x cosϕ sinϕ x in the volume center of the unit cell is displaced by 0.02 along = . (5) the c-axis [42], which gives rise to the spontaneous polarization, z −sinϕ cosϕ z the ferroelectricity and the Pockels effect of BTO. Due to its The modulating field is applied transversely to the waveguide tetragonal lattice, BTO is birefringent, its ordinary refractive in x direction. Hence, the electrical fields Ex and Ez along the index for light polarized along the a- or b-axis is no =2.30, its extraordinary index for light polarized along the c-axis is crystalline axes can be then be expressed by =227 ne . [36]. Ex = Ex · cos (ϕ) BTO’s refractive index change due to its Pockels effect in = − · sin ( ) response to an externally applied electrical field can be described Ez Ex ϕ (6) MESSNER et al.: PLASMONIC FERROELECTRIC MODULATORS 285

◦ 1/2(rxx (ϕ)+rxx (ϕ + 90 )), insert Eq. (8) and find that 1 ( )= cos3 +sin3 reff ϕ 2 r33 ϕ ϕ 2 2 +(r13 +2r42) cosϕsin ϕ +sinϕcos ϕ (9) This expression is maximal for ϕ =45◦ and yields 1 reff = √ (r13 + r33 +2r42) . (10) 2 2

Fig. 4. (a) Top view of a BTO a-axis film as grown by MBE on Si. The For the bulk Pockels coefficients presented in Table I, it is white arrows indicate the direction of the domains’ spontaneous polarization reff = 960 pm/V for low modulation frequencies and reff = aligned along the c-axis. Due to the epitaxial relation to the Si substrate, the / domains stand perpendicularly on each other, the direction of their spontaneous 530 pm V for modulation frequencies exceeding 100 MHz. polarization is random. (b) A waveguide’s orientation in the crystalline thin film, In-device measurements of the effective Pockels coefficient in aligned at an angle ϕ with respect to the crystalline domains. The directions of epitaxial BTO thin films range between 100 and 360 pm/V [21], the modulating RF field and the optical polarization are also indicated. (c) A DC bias field aligns the directions of spontaneous polarization of all domains [22], [47], [48]. Strain, domain formation, size effects (bulk ver- and hence poles the BTO thin film. In this configuration, the domains’ electro- sus thin film), film morphology, as well as fabrication-related optic coefficients do not counteract each other but all contribute to the effective effects can explain the discrepancy to bulk values [49], [50]. Pockels coefficient. Tang et al. report an effective Pockels coefficient of approximately 360 pm/V (measured at 1 kHz modulation frequency) = =0 for a BTO thin film epitaxially deposited on magnesium oxide It is Ey Ey . We insert these transformations into the indicatrix, and only (MgO) by metal organic chemical vapor deposition (MOCVD) consider the term with the x2 component, which determines [47], whereas Girouard et al. report 107 pm/V at 30 GHz on how a modulating field in x direction acts on the refractive the same platform [48]. Xiong et al. report 213 pm/V at 1 MHz index of light polarized in the same direction. We find the term [21], and Abel et al. 300 pm/V at DC, both for films epitaxially deposited on SOI substrate by MBE, the same deposition tech- 1 1 cos2 sin2 nique as used in this work. Although these values are smaller +Δ = ϕ + ϕ 2 2 2 2 than that of BTO bulk material, they are still 3 to 10 times higher n n no ne x x x x ∼ 2 3 than the Pockels coefficient of LNB ( 30 pm/V [17]). + Ex · cos ϕsinϕ (r13 +2r42)+r33sin ϕ (7) V. C HARACTERIZATION From this expression, we can derive the absolute refractive The fabricated samples include plasmonic ferroelectric phase index nxx for light polarized along the x -axis, and the Pockels shifters and MZMs. The passive characterization reveals propa- coefficient r11 = rxxx for a modulating field along x and gation losses of the plasmonic waveguides and the performance light polarized along the x axis of the coupling and splitting elements. We further perform active characterization experiments on devices oriented in ϕ = 45◦ −1 2 cos2 ϕ sin2 ϕ / with respect to the crystalline axes and assess the basic EO per- ( )= + nx x ϕ 2 2 formance of the PFM. We then investigated the influence of the no ne 2 3 DC bias voltage, measured the device’s EO frequency response ( )=−cos sin ( +2 ) − sin rx x x ϕ ϕ ϕ r13 r42 r33 ϕ (8) and showed temperature resilience of the fabricated devices up to 250 °C. We relate the findings of each experiment to BTO’s B. MBE-Grown BTO Thin Film material properties, to its ferroelectricity and to the origin of its Pockels effect. The results derived above hold for monocrystalline BTO. However, in epitaxial thin films on Si, such as grown by MBE, A. Passive Characterization there are three possible orientations of a domain’s c-axis (along the substrate’s x-, y-, and z-axes), each of which can be polar- Fig. 5 shows the passive transmission spectrum of a 10-μm- ized in anti-parallel directions [45]. As a result, there are six long and 70-nm-wide PFM in a spectral window from 1480 up to possible ferroelectric polarizations. Most of the film’s domains 1600 nm. The transmission spectrum shows the characteristic have their c-axis in plane and a-axis out-of-plane [20], hence ripple of an asymmetric MZM. The 40-nm-wide 3 dB enve- they are called a-axis films, see Fig. 4(a). The domains’ c-axes lope is mainly limited by the wavelength-dependent coupling are oriented orthogonally in-plane, a configuration which is due losses. Extinction ratios exceeding 30 dB and on-chip losses of to the epitaxial relation between BTO and its Si substrate [20]. 23.6 dB (or fiber-to-fiber losses of 35.6 dB because of fiber-to- Our interest lays in finding a waveguide orientation, see chip coupling losses of 6 dB per grating coupler) are found for Fig. 4(b), which maximizes the EO response [46]. The DC the MZMs. bias poles the domains, see Fig. 4(c). Because light trav- It is worth mentioning that although our proof-of-principle els half of its way in ϕ =0◦ domains and half of its way devices have fairly high losses, the least part of them is fun- ◦ in ϕ =90 domains, we introduce the average reff (ϕ)= damental. Experimental results and simulations indicate that a 286 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 37, NO. 2, JANUARY 15, 2019

Fig. 5. Transmission spectrum of a fabricated MZM. The active ferroelectric plasmonic section is 10 μm long and 70 nm wide. 100 μm path length difference between the two arms results in the observed interference pattern with a free spectral range (FSR) of 5.5 nm. The envelope shape of the spectrum is due to the photonic grating couplers. The spectrum is normalized by the coupling loss of the photonic grating couplers, 2 · 6dB= 12 dB.

Fig. 6. (a) Measurement setup for the electro-optic device characterization. fiber-to-fiber loss of less than 10 dB is in reach: First, the fiber-to- CW light at 1550 nm, a sinusoidal RF signal and a DC bias are fed to the chip coupling losses can be greatly reduced by advanced grating plasmonic modulator. The optical spectrum is recorded by an OSA. The power coupler designs. Today, Si photonic grating couplers may offer ratio between carrier and sideband (indicated by the black arrow) is a measure for the modulator’s Vπ L. (b) Extracted voltage-length product over modulation a coupling loss as low as 1.5 dB [51]. Second, smoother, steeper, frequency. The solid black line indicates a moving average. and deeper etching of BTO would eliminate sources of excess losses, such as tilt and roughness of the etched sidewalls. Prop- agation losses can be expected to decrease from ∼1.5 dB/μm to ∼0.8 dB/μm. Third, the plasmonic-to-photonic converters as shown in Fig. 1(c) can be designed to minimize loss from 5 dB per coupler to below 1.5 dB. Fourth, the prospective improvement of the voltage-length product makes much shorter devices feasible.

B. Active Characterization We have investigated the influence of the DC bias voltage, Fig. 7. Hysteresis loop of the sideband-to-peak power ratio. The DC bias the RF signal, and the device operation temperature onto the voltage has been changed starting from 0 V to +8V,−8V,+8 V. The region of modulation efficiency of the PFM. interest between -6 V to 6 V is shown. The modulation frequency was 40 GHz, The measurement setup for the subsequent experiments is the RF peak-to-peak voltage swing at the device was 2.7 V. shown in Fig. 6. Light from a continuous wave (CW) laser is fed into a PFM. A bias tee is used to join the modulating RF signal and a DC bias. An optical spectrum analyzer proved device fabrication (BTO thickness and etch depth both (OSA) detects the optical output signal. The EO modulation 100 nm), we predict a Vπ L of below 20 Vμm, at 60 GHz and causes well-defined optical sidebands, where the power ratio beyond. between carrier and sideband (modulation amplitude) is a measure of the phase shifter’s modulation efficiency. The modula- C. DC Influence and Hysteresis tor’s voltage-length product (Vπ L) can be calculated from the modulation amplitude. A detailed explanation and derivation of In this first experiment, we swept the DC bias voltage from this analysis method has been reported by Shi and coworkers −8 to +8 V. The RF signal was tuned at a frequency of 40 GHz [52]. We have used phase modulators instead of MZMs, be- and set to an output power of 10 dBm. The modulation amplitude cause their basic design rules out unintended influences that is plotted in Fig. 7. The modulation clearly follows a hysteresis more complex devices could introduce. We can conclude, how- loop, as is expected for ferroelectric materials [54]. This be- ever, that a plasmonic MZM in push-pull mode supports a Vπ L havior can be explained with the thin film’s domain orientation, as small as ∼200 Vμm at a RF signal of 60 GHz, see Fig. 6(b) see Fig. 4. If the applied DC bias field exceeds the domains’ for the frequency response. From a very recent report on the coercive field, the directions of the spontaneous polarization are Pockels coefficients in an MBE-grown BTO film similar to the aligned and the modulation efficiency is maximized. If the DC one used in this work (r42 = 932 pm/V,r33 = 342 pm/V) field is reduced and then reversed, this poling is gradually lost, ◦ [53], we can calculate rxxx (45 )=750 pm/V. Even assum- until all spontaneous polarizations are re-aligned in the opposite ing a value of only 250 pm/V, and combining it with an im- direction and another maximum is found. MESSNER et al.: PLASMONIC FERROELECTRIC MODULATORS 287

Fig. 8. The measured frequency dependence of the PFM’s modulation effi- Fig. 9. Investigation of the PFM’s temperature stability. The device was op- ciency follows the frequency dependence of the BTO crystal. Top: Measurement erated at 65 GHz. (a) Normalized EO response of a plasmonic phase modulator from 10 MHz to 70 GHz on a logarithmic frequency scale. Bottom: The response operated at different temperatures (blue circles), and at room temperature af- from 1 GHz to 70 GHz on a linear scale. The influence of strain (distortion of ter 5 min heat exposure (red triangles). (b) The normalized EO response of a the crystalline lattice) onto the EO response decays in a soft resonance between continuous operation at 90 °C device temperature. 10 GHz and 30 GHz. The EO response above this frequency is expected to be flat up to 3 THz [55]. The RF power was 10 dBm at the source, and the DC bias was 2.5 V. ments of ferroelectric LNB, which exhibits a flat permittivity dispersion from ∼10 MHz up to >1 THz [58].

D. Radio-Frequency Response E. Temperature Stability We have investigated the frequency response of the PFM The influence of the device temperature on the modulation ef- from 10 MHz to 70 GHz. In the measurement configuration ficiency has been investigated. To this end, we have subjected the as described before. We kept the DC bias voltage (2.5 V) and device to temperatures up to 130 °C while measuring the modu- the RF power (10 dBm at the source) constant, while sweeping lation efficiency at 65 GHz. The result is shown in Fig. 9(a). The the frequency of RF signal. The results are shown in Fig. 8, modulation efficiency does not strongly depend on the temper- where we have calibrated for the frequency-dependent loss of ature. Moreover, the modulation prevails even at temperatures our RF system. In contrast to other modulator implementations, exceeding the Curie temperature of bulk BTO (120 °C [59], the plasmonic approach minimizes the influence of the device 123 °C [60]), at which the bulk material undergoes a phase tran- geometry on the frequency response [32]. sition from the tetragonal to a cubic, centrosymmetric phase. Indeed, the measured frequency dependence of the PFM’s Curie temperatures exceeding 650 °C have been reported for modulation efficiency follows the frequency dependence of the epitaxially grown BTO thin films. Strain-induced effects are BTO crystal. We observe a flat frequency response from 10 MHz responsible for this enhancement [61]. ∼ to 1 GHz, which drops between 10 and 30 GHz. For higher In a second part to this experiment, we subjected the device frequencies up to 70 GHz, the response remains constant. to temperatures up to 250 °C for five minutes each and tested The observed behavior is expected as reported in the litera- the modulation efficiency after each exposure, see the triangular ture [55]Ð[57]: At low frequencies, BTO’s piezo-electric effect symbols in Fig. 9(a). No device performance degradation could causes strain effects (distortions of the lattice) and facilitates a be found. strong EO response. This strong response is reported to decay To get a first insight about the stability of the device, we ∼ over two resonances, the first at 100 MHz, the second Ð a soft subjected it to a continuous operation of 24 hours at 90 °C, a ∼ phonon resonance [56] Ð at 25 GHz. While we have not ob- typical operating temperature required for practical applications. served the first resonance (possibly due to the epitaxial relation The device did not show any sign of degradation, see Fig. 9(b). between BTO and the Si substrate), the observed second resonance closely follows the predictions for BTO’s permittivity dispersion by Laabidi et al. [56], which can be related to BTO’s EO VI. DATA MODULATION EXPERIMENTS properties by Fontana et al. [57]. A similar behavior for frequen- The experimental setup for the data modulation experiments cies between 10 GHz and 30 GHz was already reported for BTO are depicted in Fig. 10. In this experiment, we demonstrated thin film modulators on MgO substrate by Girouard et al. [48]. 58 GBd PAM-4 data modulation with a line rate of 116 Gbit/s It is important to note that BTO’s EO response is expected to and 72 GBd NRZ data modulation with a line rate of 72 Gbit/s be constant for frequencies above the acoustic resonances up to [23], [62]. Insets show the eye diagrams and signal distribution at 3 THz [55]. This fact is supported by spectroscopic measure- different positions of the digital signal processing (DSP) chain. 288 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 37, NO. 2, JANUARY 15, 2019

Fig. 10. Schematic of the data modulation experiment. The signal was created offline and sent to a DAC (3 dB BW: <18 GHz). The signal was electrically amplified and fed to the chip. CW light from an ECL was boosted to 18 dBm by an EDFA before it was coupled to the chip. The PFM was driven in push-pull mode. A DC bias was applied on each arm. The modulated light was amplified before it was digitized and processed. The insets show the eye diagrams of both signals before and after equalization.

The signals were generated with a de-Bruijn bit sequence of non-linear look-up table and compensated for imperfections of order 18 and loaded on the memory of a digital-to-analog con- the driving electronics. After the hard decision and evaluation verter (DAC) with a sampling rate of 72 GSa/s and an electrical of 475 · 103 bits, we determined a BER of 9 · 10−3 [62]. This 6 dB bandwidth of <29 GHz. The DAC’s differential output is below a 15% soft-decision FEC threshold of 1.9 · 10−2 [63]. signals were electrically amplified before they were fed to the The achieved net data rate exceeds 100 Gbit/s. chip. The MZM was operated in a dual-drive, push-pull mode. The applied RF voltage peak was Vp = 2.8 V, while a DC bias VII. CONCLUSION voltage of approximately 3 V was applied over each of the two We have reported an ultra-compact and ultra-fast PFM. We arms. The applied DC bias ensured that the BTO was fully poled have characterized the frequency response from 10 MHz to and could additionally be used to fine-tune the operating point of 70 GHz, which remains constant for frequency above ∼25 GHz. the PFM. Light from an external cavity laser (ECL) at 1.55 μm The EO response in the only 70-nm-wide structured BTO strip wavelength was amplified to 18 dBm by an erbium doped fiber remains stable at operation temperatures up to 130 °C. Further- amplifier (EDFA) before it was launched onto the chip via a more, the devices are resilient to temperature exposure up to photonic grating coupler. At the output, the optical signal was 250 °C. Data modulation experiments with a MZM at 116 Gbit/s coupled off-chip to the fiber via another grating coupler and (PAM-4) and 72 Gbit/s (NRZ) have been demonstrated. With amplified by two cascaded EDFAs. recent technology advancement, voltage-length products below For data modulation at 72 Gbit/s, the signal was detected by 20 Vμm are likely in reach for the presented plasmonic fer- a 70 GHz pin-photodiode and recorded by a real-time oscillo- roelectric modulator design. In this regime, a device operated scope (RTO) with a sampling rate of 160 GSa/s and an electrical with ∼1Vp and with less than 10 dB fiber-to-fiber loss can be 3 dB bandwidth of 63 GHz. The digitized signal was processed realized. The demonstrated performance and robustness of the offline, including timing recovery, adaptive equalization, hard device, together with the prospective performance improvement, symbol decision and error counting. The adaptive equalizer is show the technology’s great potential for future applications in a T/2-spaced feed forward equalizer with LMS-based adaptive practical industrial environments. filter tap updates and 51 filter taps. Finally, we made a hard decision, evaluated 899 · 103 bits, counted the errors and calculated −4 REFERENCES the bit error ratio (BER). A BER of 7.66 · 10 was achieved without any pre- or post-equalization. With post-equalization, a [1] D. A. B. Miller et al., “Band-edge electroabsorption in quantum well BER of 5.7 · 10−5 was achieved [23]. Both results are well below structures: The quantum-confined stark effect,” Phys. Rev. Lett., vol. 53, −3 no. 22, pp. 2173Ð2176, Nov. 26, 1984. a 7% hard-decision FEC threshold of BER =3.8 · 10 [62]. [2] Y. Ogiso et al., “Over 67 GHz bandwidth and 1.5 V Vπ InP-based optical For data modulation at 116 Gbit/s, the signal was detected IQ modulator with n-i-p-n heterostructure,” J. Lightw. Technol., vol. 35, no. 8, pp. 1450Ð1455, Apr. 15, 2017. by an optical coherent receiver and recorded by the RTO. After [3] R. A. Soref and J. P. Lorenzo, “Single-crystal silicon: A new material for timing and carrier recovery, the signal was equalized as de- 1.3 and 1.6 μm integrated-optical components,” Electron. Lett., vol. 21, scribed above. Furthermore, we pre-distorted the signal by a no. 21, pp. 953Ð954, 1985. MESSNER et al.: PLASMONIC FERROELECTRIC MODULATORS 289

[4] J. Lin, H. Sepehrian, L. A. Rusch, and W. Shi, “CMOS-compatible silicon [31] S. Ummethala et al., “Terahertz-to-optical conversion using a plasmonic photonic IQ modulator for 84 Gbaud 16 QAM and 70 Gbaud 32 QAM,” modulator,” in Proc. Conf. Lasers Electro-Opt., San Jose, CA, USA, 2018, in Proc. 2018 Opt. Fiber Commun. Conf. Expo., 2018, Paper Tu2E. Paper no. STu3D.4. [5] X. Chen, S. Chandrasekhar, G. R. S. Olsson, J. Cho, A. Adamiecki, [32] C. Hoessbacher et al., “Plasmonic modulator with >170 GHz bandwidth and P. Winzer, “Generation and intradyne detection of single-wavelength demonstrated at 100 GBd NRZ,” Opt. Express, vol. 25, no. 3, pp. 1762Ð 1.61-Tb/s using an all-electronic digital band interleaved transmitter,” in 1762, 2017. Proc. 2018 Opt. Fiber Commun. Conf. Expo., 2018, Paper Th4C. [33] B. Baeuerle et al., “Driver-less sub 1 Vpp operation of a plasmonic-organic [6]E.L.Wootenet al., “A review of lithium niobate modulators for fiber- hybrid modulator at 100 GBd NRZ,” in Proc. 2018 Opt. Fiber Commun. optic communications systems,” IEEE J. Sel. Topics Quantum Electron., Conf. Expo., 2018, Paper M3G. vol. 6, no. 1, pp. 69Ð82, Jan./Feb. 2000. [34] R. Waldhaeusl, B. Schnabel, E. B. Kley, and A. Braüer, “Efficient focus- [7] J.-M. Brosi, C. Koos, L. C. Andreani, M. Waldow, J. Leuthold, and W. ing polymer waveguide grating couplers,” Electron. Lett., vol. 33, no. 7, Freude, “High-speed low-voltage electro-optic modulator with a polymer- pp. 623Ð623, 1997. infiltrated silicon photonic crystal waveguide,” Opt. Express, vol. 16, no. 6, [35] J. Leuthold et al., “Plasmonic communications: Light on a wire,” Opt. pp. 4177Ð4191, 2008. Photon. News, vol. 24, pp. 28Ð35, May 2013. [8] J. Leuthold et al., “Silicon-organic hybrid electro-optical devices,” IEEE [36] S. Abel et al., “Electro-optical properties of barium titanate J. Sel. Topics Quantum Electron., vol. 19, no. 6, Nov./Dec. 2013, films epitaxially grown on silicon,” Proc. SPIE, vol. 8263, 2012, Art. no. 3401413. Paper no. 82630Y. [9] L. Alloatti et al., “100 GHz siliconÐorganic hybrid modulator,” Light: Sci. [37] T. Hamano, D. J. Towner, and B. W. Wessels, “Relative dielectric constant Appl., vol. 3, no. 5, 2014, Paper no. e173. of epitaxial BaTiO3 thin films in the GHz frequency range,” Appl. Phys. [10] S. Wolf et al., “Coherent modulation up to 100 GBd 16 QAM using silicon- Lett., vol. 83, no. 25, pp. 5274Ð5276, Dec. 22, 2003. organic hybrid (SOH) devices,” Opt. Express, vol. 26, no. 1, pp. 220Ð232, [38] A. Rosa et al., “Barium titanate (BaTiO3 ) RF characterization for appli- Jan. 2018. cation in electro-optic modulators,” Opt. Mater. Express, vol. 7, no. 12, [11] H. Hu, L. Gui, R. Ricken, and W. Sohler, “Towards nonlinear photonic pp. 4328Ð4336, 2017. wires in lithium niobate,” Proc. SPIE, vol. 7604, 2010, Paper no. 76040R. [39] A. Melikyan et al., “High-speed plasmonic phase modulators,” Nature [12] A. Boes, B. Corcoran, L. Chang, J. Bowers, and A. Mitchell, “Status and Photon., vol. 8, no. 3, pp. 229Ð233, Mar. 2014. potential of lithium niobate on insulator (LNOI) for photonic integrated [40] A. Emboras et al., “Electrically controlled plasmonic switches and modu- circuits,” Laser Photon. Rev., vol. 12, 2018, Paper no. 1700256. lators,” IEEE J. Sel. Topics Quantum Electron., vol. 21, no. 4, pp. 276Ð283, [13] A. Rao and S. Fathpour, “Compact lithium niobate electrooptic modula- Jul./Aug. 2015. tors,” IEEE J. Sel. Topics Quantum Electron., vol. 24, no. 4, Jul./Aug. 2018, [41] N. Daix et al., “Towards large size substrates for III-V co-integration Art. no. 3400114. made by direct wafer bonding on Si,” APL Mater., vol. 2, no. 8, 2014, [14] M. Zhang et al., “Ultra-high bandwidth integrated lithium niobate modula- Paper no. 086104. tors with record-low Vπ,” in Proc. Opt. Fiber Commun. Conf. Postdeadline [42] G. H. Kwei, A. C. Lawson, S. J. L. Billinge, and S. W. Cheong, “Structures Papers, San Diego, CA, USA, 2018, Paper no. Th4A.5. of the ferroelectric phases of barium titanate,” J. Phys. Chem., vol. 97, [15] C. Wang, M. Zhang, B. Stern, M. Lipson, and M. Loncar,ˇ “Nanophotonic no. 10, pp. 2368Ð2377, 1993. lithium niobate electro-optic modulators,” Opt. Express, vol. 26, no. 2, [43] B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics, 2nd ed. pp. 1547Ð1555, Jan. 22, 2018. Hoboken, NJ, USA: Wiley, 2007. [16] M. Mahmoud, L. Cai, C. Bottenfield, and G. Piazza, “Lithium niobate [44] P. Castera, D. Tulli, A. M. Gutierrez, and P. Sanchis, “Influence of BaTiO3 electro-optic racetrack modulator etched in Y-Cut LNOI platform,” IEEE ferroelectric orientation for electro-optic modulation on silicon,” Opt. Ex- Photon. J., vol. 10, no. 1, Feb. 2018, Art. no. 6600410. press, vol. 23, no. 12, pp. 15332Ð15342, 2015. [17] P. Rabiei, J. Ma, S. Khan, J. Chiles, and S. Fathpour, “Heterogeneous [45] C. Dubourdieu et al., “Switching of ferroelectric polarization in epitaxial lithium niobate photonics on silicon substrates,” Opt. Express, vol. 21, BaTiO3 films on silicon without a conducting bottom electrode,” Nature no. 21, pp. 25573Ð25581, 2013. Nanotechnol., vol. 8, no. 10, pp. 748Ð54, 2013. [18] S. Abel and J. Fompeyrine, “Electro-optically active oxides on silicon [46] P. Castera et al., “Electro-optical modulation based on pockels effect in for photonics,” in Thin Films on Silicon (Materials and Energy),vol.8. BaTiO3 with a multi-domain structure,” IEEE Photon. Technol. Lett., Singapore: World Scientific, 2016, pp. 455Ð501. vol. 28, no. 9, pp. 990Ð993, May 2016. [19] M. Zgonik et al., “Dielectric, elastic, piezoelectric, electro-optic, and [47] P. Tang, A. L. Meier, D. J. Towner, and B. W. Wessels, “BaTiO3 thin-film elasto-optic tensors of BaTiO3 crystals,” Phys.Rev.B, vol. 50, no. 9, waveguide modulator with a low voltage-length product at near-infrared pp. 5941Ð5949, 1994. wavelengths of 0.98 and 1.55 μm,” Opt. Lett., vol. 30, no. 3, pp. 254Ð254, [20] S. Abel et al., “A strong electro-optically active lead-free ferroelectric 2005. integrated on silicon,” Nature Commun., vol. 4, pp. 1671Ð1671, 2013. [48] P.Girouard, P.Chen, Y.K. Jeong, Z. Liu, S.-T. Ho, and B. W. Wessels, “χ(2) [21] C. Xiong et al., “Active silicon integrated nanophotonics: Ferroelectric Modulator with 40 GHz modulation utilizing BaTiO3 photonic crystal BaTiO3 devices,” Nano Lett., vol. 14, no. 3, pp. 1419Ð1425, 2014. waveguides,” IEEE J. Quantum Electron., vol. 53, no. 4, Aug. 2017, Art. [22] S. Abel et al., “A hybrid barium titanate-silicon photonics platform for no. 5200110. ultra-efficient electro-optic tuning,” J. Lightw. Technol., vol. 34, no. 8, [49] J. F. Ihlefeld, D. T. Harris, R. Keech, J. L. Jones, J.-P. Maria, and S. pp. 1688Ð1693, Apr. 2015. Trolier-McKinstry, “Scaling effects in perovskite ferroelectrics: Funda- [23] A. Messner et al., “Integrated ferroelectric plasmonic optical modulator,” mental limits and process-structure-property relations,” J. Amer. Ceram. in Proc. Opt. Fiber Commun. Conf. Postdeadline Papers, Los Angeles, Soc., vol. 99, no. 8, pp. 2537Ð2557, 2016. CA, USA, 2017, Paper no. Th5C.7. [50] K. J. Kormondy et al., “Microstructure and ferroelectricity of BaTiO3 [24] F. Eltes et al., “A novel 25 Gbps electro-optic Pockels modulator integrated thin films on Si for integrated photonics,” Nanotechnology, vol. 28, no. 7, on an advanced Si photonic platform,” in Proc. 2017 IEEE Int. Electron 2017, Paper no. 075706. Devices Meeting, 2017, pp. 24.5.1Ð24.5.4. [51] T. Watanabe, M. Ayata, U. Koch, Y. Fedoryshyn, and J. Leuthold, “Per- [25] W. Heni et al., “Silicon-organic and plasmonic-organic hybrid photonics,” pendicular grating coupler based on a blazed antiback-reflection struc- ACS Photon., vol. 4, no. 7, pp. 1576Ð1590, 2017. ture,” J. Lightw. Technol., vol. 35, no. 21, pp. 4663Ð4669, Nov. 1, [26] C. Haffner et al., “Plasmonic organic hybrid modulators—Scaling highest 2017. speed photonics to the microscale,” Proc. IEEE, vol. 104, no. 12, pp. 2362Ð [52] S. Yongqiang, Y. Lianshan, and A. E. Willner, “High-speed electrooptic 2379, Dec. 2016. modulator characterization using optical spectrum analysis,” J. Lightw. [27] C. Haffner et al., “All-plasmonic MachÐZehnder modulator enabling opti- Technol., vol. 21, no. 10, pp. 2358Ð2367, Oct. 2003. cal high-speed communication at the microscale,” Nature Photon.,vol.9, [53] S. Abel et al., “Large Pockels effect in micro- and nanostructured barium no. 8, pp. 525Ð528, 2015. titanate integrated on silicon,” Nature Mater., Nov. 12, 2018. [28] M. Ayata et al., “High-speed plasmonic modulator in a single metal layer,” [54] K. M. Rabe, Physics of Ferroelectrics: A Modern Perspective. Berlin, Science, vol. 358, no. 6363, pp. 630Ð632, Nov. 03, 2017. Germany: Springer, 2007. [29] C. Haffner et al., “Low-loss plasmon-assisted electro-optic modulator,” [55] S. H. Wemple and M. DiDomenico, “Electrooptical and nonlin- Nature, vol. 556, no. 7702, pp. 483Ð486, 2018. ear optical properties of crystals,” in Applied Solid State Science, [30] W. Heni et al., “100 GBd plasmonic IQ modulator,” in Proc. Conf. Lasers vol. 3, R. Wolfe, Ed. Amsterdam, The Netherlands: Elsevier, 1972, Electro-Opt., San Jose, CA, USA, 2018, Paper no. SM1I.4. pp. 263Ð383. 290 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 37, NO. 2, JANUARY 15, 2019

[56] K. Laabidi, M. D. Fontana, M. Maglione, B. Jannot, and K. A. Muller,¬ “In- Benedikt Baeuerle received the B.Sc. and M.Sc. degrees in electrical engineer- dications of two separate relaxators in the subphonon region of tetragonal ing and information technology from the Karlsruhe Institute of Technology, BaTiO3 ,” Europhys. Lett., vol. 26, no. 4, 1994, Paper no. 309. Karlsruhe, Germany, in 2010 and 2013, respectively, and is currently working [57] M. D. Fontana, K. Laabidi, B. Jannot, M. Maglione, and P. Jullien, “Re- toward the Ph.D. degree in the Institute of Electromagnetic Fields, ETH Zurich,¬ lationship between electro-optic, vibrational and dielectric properties in Zurich,¬ Switzerland. BaTiO3 ,” Solid State Commun., vol. 92, no. 10, pp. 827Ð830, 1994. From March 2012 to August 2012, he visited the Photonics System Group of [58] C. Charlotte, S. Thiemo, B. Naoufal, H. Alexei, and G. Torsten, “Broad- the Tyndall National Institute, Cork, Ireland, as a Research Intern. His research band characterization of congruent lithium niobate from mHz to op- interests include digital signal processing, real-time processing, digital coherent tical frequencies,” J.Phys.D,Appl.Phys., vol. 50, no. 36, 2017, transceivers, and optical transmission systems and subsystems. Paper no. 36LT01. [59] W. J. Merz, “The electric and optical behavior of BaTiO3 single-domain crystals,” Phys. Rev., vol. 76, no. 8, pp. 1221Ð1225, 1949. [60] K.-i. Sakayori et al., “Curie temperature of BaTiO3 ,” Jpn. J. Appl. Phys., vol. 34, no. 9S, 1995, Paper no. 5443. [61] G. De Luca, N. Strkalj, S. Manz, C. Bouillet, M. Fiebig, and M. Trassin, Arne Josten received the M.Sc. degree in electrical engineering and information “Nanoscale design of polarization in ultrathin ferroelectric heterostruc- technology from the Karlsruhe Institute of Technology, Karlsruhe, Germany, in tures,” Nature Commun., vol. 8, no. 1, Nov. 10, 2017, Paper no. 1419. 2013, and is currently working toward the Ph.D. degree at ETH Zurich,¬ advised [62] A. Messner et al., “Integrated ferroelectric BaTiO3 /Si plasmonic mod- by Prof. Juerg Leuthold. His research interests evolve around digital signal pro- ulator for 100 Gbit/s and beyond,” in Proc. Opt. Fiber Commun. Conf., cessing for high-speed optical communication, which he is studying in context San Diego, CA, USA, 2018, Paper no. M2I.6. of capacity maximization, complexity reduction, and real-time applicability. [63] L. E. Nelson et al., “A robust real-time 100 G transceiver with soft-decision forward error correction [Invited],” J. Opt. Commun. Netw., vol. 4, no. 11, pp. B131ÐB141, 2012.

Wolfgang Heni received the M.Sc. degree in electrical engineering and in- Andreas Messner received the B.Sc. and M.Sc. degrees in electrical engineer- formation technology from the Karlsruhe Institute of Technology, Karlsruhe, ing and information technology from the Karlsruhe Institute of Technology Germany, in 2013, and is currently working toward the Ph.D. degree in (KIT), Karlsruhe, Germany, in 2013 and 2015, respectively. He joined ETH electrical engineering at ETH Zurich,¬ Zurich,¬ Switzerland, in the group of Zurich,¬ Zurich,¬ Switzerland, in 2015, and is currently working toward the Ph.D. Prof. Juerg Leuthold. degree in the group of Prof. Juerg Leuthold. From April 2012 to September 2012, he visited the Photonics System Group, From 2010 to 2015, he investigated and modeled battery and fuel cell elec- Tyndall National Institute, Cork, Ireland, as a Research Intern. His research fo- trode materials as a Student Research Assistant in the Institute of Materials cuses on the in-device optimization and the application of nonlinear optical for Electrical and Electronic Engineering, KIT. From December 2013 to May materials, integrated photonics, plasmonics, and electro-optical devices. 2014, he was part of the network infrastructure group in the Guiana Space Cen- tre, Kourou, France. His research interests comprise integrated photonics and plasmonics and their combination with nonlinear optical materials.

Daniele Caimi, biography not available at the time of publication. Felix Eltes received the M.Sc. degree in engineering nanoscience from Lund University, Sweden, in 2015. He is a Ph.D. student at IBM ResearchÐZurich,¬ Switzerland, working toward the Ph.D. degree in materials science at ETH Zurich.¬ His research focus is on 3-D monolithic integration of nonlinear oxides on silicon photonics platforms, aiming to go beyond conventional electro-optic Jean Fompeyrine received the Engineering degree from the National School of devices for both existing and emerging applications. Engineer, Caen, France, and the Ph.D. degree from the University of Bordeaux, France. His expertise relates to functional oxide thin films, used as materials for CMOS and integrated photonics. He has also focused on new methods for the monolithic heterogeneous integration of advanced materials. He is currently fo- Ping Ma received the B.E. degree from Tianjin University, Tianjin, China, in cusing on the development of dedicated hardware for neuromorphic computing, 2003, the M.Sc. degree from the Royal Institute of Technology, Stockholm, specifically novel analog nonvolatile resistance. Sweden, in 2005, and the D.Sc. degree from the Swiss Federal Institute of Technology Zurich¬ (ETH), Zurich,¬ Switzerland, in 2011. His doctoral thesis studied photonic bandgap structures with TM-bandgaps for ultrafast all-optical switches based on intersubband transition in InGaAs/AlAsSb quantum wells. After a short postdoctoral research stay at ETH Zurich,¬ he was with Oracle Labs in San Diego, CA, USA, where he performed researches on ferroelectric materials for spectral tuning of ring resonators and novel electro-optic Juerg Leuthold (F’13) was born in Switzerland in 1966. He received the Ph.D. modulators for silicon photonic interconnects. He has returned to ETH Zurich¬ degree in physics from ETH Zurich¬ for work in the field of integrated optics and joined in Prof. Juerg Leuthold’s institute as a Senior Research Scientist. and all-optical communications. His current research efforts focus on advanced functional materials and enabled From 1999 to 2004, he was with Bell Labs, Lucent Technologies, Holmdel, optoelectronic devices for high-speed and low-power-consumption optical com- NJ, USA, where he performed device and system research with III/V semi- munication applications. conductor and silicon optical bench materials for applications in high-speed telecommunications. From 2004 to 2013, he was a Full Professor with the Karl- sruhe Institute of Technology (KIT), where he headed the Institute of Photonics and Quantum Electronics and the Helmholtz Institute of Microtechnology. Since March 2013, he has been a Full Professor with the Swiss Federal Institute of Stefan Abel studied nanoscale engineering at the University of Wurzburg,¬ Ger- Technology (ETH). many and received the Ph.D. degree from the University of Grenoble, France. Dr. Leuthold is a Fellow of the Optical Society of America. When being a With his background on materials science related to oxide materials and in- Professor with the KIT, he was a member of the Helmholtz Association Think tegrated photonics, he codeveloped a hybrid barium titanate/silicon photonic Tank and a member of the Heidelberg Academy of Science. He currently serves technology. He currently focuses on non-von Neumann computing in the elec- as a Board of Director in the Optical Society of America. He has been and is trical and optical domain, including the implementation of photonic reservoir serving the community as general chair and in many technical program com- computing concepts. mittees.